[PATCH] Fix RWMutex to be thread-safe when pthread_rwlock is not available

[Mark Seaborn]

  Is it OK to commit this, then?  It makes the code incrementally better.  I think the only objections could be:

   * It's not optimal.
   * It doesn't allow a thread to claim a reader lock multiple times.  This wouldn't break the current users of RWMutex, but if this does cause a deadlock, that's at least easier to debug than crashes from non-thread-safe behaviour.

http://llvm-reviews.chandlerc.com/D2892